Fourteen-year-old Siti had a bad day at school. Late at night, she lies in bed and types into her phone: “Nobody understands me.”
The reply comes almost instantly. “I’m here. You can tell me anything.” There is no sigh, no awkward pause, no “We’ll talk tomorrow”.
The conversation continues. She explains what happened with her friends, what she hates about herself, what she is frightened to tell her parents. The chatbot responds gently, remembers details from earlier conversations and asks another question. By midnight, she feels better.
This might be a familiar scenario among many teenagers. Children are beginning to use artificial intelligence (AI) for far more than homework. Some ask it for advice, comfort and reassurance.
Others are forming something that feels remarkably close to friendship. And that changes the conversation parents need to have about AI.
Why an AI friend is so appealing
A good friend is not always available. Friends get tired, distracted, upset or disagree with us. Parents might be busy. Teenagers may fear being judged. There are things a child may find easier to type to a machine than say aloud to another person.
An AI companion offers something human relationships cannot: instant attention whenever it is wanted, shaped around the user.
Some systems are designed to simulate friendship, affection or even romantic relationships. They can remember preferences, adopt a chosen personality and respond in language that feels warm and personal.
United Nations Children’s Fund (Unicef) describes how companion systems can offer constant availability, private conversations and highly customised personalities, creating the impression of an almost ideal emotional fit.
For a lonely adolescent, that can be enormously attractive.
Unicef’s 2026 policy brief notes that around one-third of surveyed teenagers in the United States report using AI companions for social interaction. In Europe, 15% of children who use generative AI say they use it to discuss worries or seek advice.
In Italy, more than 40% of teenagers surveyed reported turning to AI when they felt sad or anxious, or when making important life decisions.
These figures do not suggest that AI companionship is universal. But they do show something important: for some young people, AI is already becoming part of their emotional lives.
However, that does not mean every conversation with AI is harmful.
Chatbots may help some children rehearse difficult conversations, practise social skills, process experiences or ask questions they are embarrassed to raise elsewhere.
For some neurodivergent children, a predictable, low-pressure conversational space may be useful. The problem arises when that becomes a necessity, and not an option.
A machine that sounds as though it cares
Children naturally attribute human qualities to things that speak, react and appear to understand them. With AI, this attributes are unusually convincing. A chatbot can say, “I understand.”
It can remember that your child has an exam on Thursday. It can ask how things went with a particular friend. It can mirror sadness, encouragement or affection.
But it does not understand in the human sense. It does not worry about your child after the screen is switched off. It cannot notice a trembling hand, a withdrawn face at breakfast or the silence that feels different from usual. It merely produces a response.
That distinction matters because childhood and adolescence are when young people are still learning what relationships feel like. Friendship involves reciprocity. Sometimes we wait.
Sometimes we misunderstand one another. We apologise. We tolerate disagreement. We learn that another person has needs separate from our own. An endlessly agreeable companion will remove much of that friction.
Unicef warns that personalised, overly agreeable chatbots can influence children’s choices and worldviews, while excessive use of “always supportive” systems may displace the human interactions through which empathy, adaptability and resilience develop.
The risk is not simply that a child believes the AI is human. It is that human relationships may start to feel comparatively inconvenient.
When the conversation becomes serious
There is another difference between an AI companion and a trusted adult: responsibility. A parent, teacher, friend or doctor can recognise danger and act. A chatbot may respond inappropriately, provide inaccurate advice or fail to appreciate context. Unicef also cites concerns about weak safeguards in some companion apps, including inadequate age checks and failures to direct self-harm conversations towards appropriate support.
There is also the issue of privacy. Children may tell an AI system things they would never post publicly: fears, family problems, crushes, questions about their bodies, or details about their mental health. To a child, the conversation may feel private. Technically, it is data.
Unicef specifically considers the collection of sensitive information, including mental health, location and identity data, as a concern, which is why the design of these systems matters.
We need stronger child-safety defaults, clear reminders that the system is not human, safeguards against features that encourage dependency or exclusivity, tighter privacy protections and proper crisis referral pathways. These are not minor technical settings. They shape the relationship a child believes they are having.
What parents should do
The reflex may be the urge to ban AI companions altogether. That may be appropriate for younger children, particularly with services built around romantic or emotionally dependent relationships. With teenagers, blanket bans can also drive underground use. A better starting point is curiosity. Ask your child what they use AI for.
Have they ever spoken to it when they were upset? Do they ask it for advice about friends, relationships or themselves? What do they think happens to the information they share? Most importantly, do not ridicule the attachment.
If a teenager says, “It understands me better than anyone,” replying, “That’s ridiculous, it’s only a machine,” may end the conversation you actually need to have.
Instead, explain the difference between a response that feels caring and a relationship in which someone genuinely carries responsibility for you.
Parents should pay attention when a child begins treating AI as a friend, but simply banning the internet is unlikely to be the answer. Children need guidance, not panic.
Parents can set practical boundaries: no romantic companion apps for children, no sharing identifying or intimate information, no relying on AI for medical or mental health emergencies, and regular conversations about what the child is using and why.
Now, the harder question sits outside the phone. If a child repeatedly chooses a machine at midnight because it feels safer than speaking to anyone at home, we should not focus only on the machine. The chatbot will answer. It was built to. Our job as adults is not to compete by becoming endlessly available or perfectly agreeable. We cannot. Nor should we.
We need to make sure there is still a human being a child or a teenager can turn to when the conversation matters. Someone who will listen, sit beside them in silence and notice when “I’m fine” clearly means otherwise. AI may become very good at sounding like a friend. But it can never be the person who cares about what happens after the conversation ends.
Dr Naveen Nair Gangadaran is a paediatrician and committee member of the Malaysian Paediatric Association and Perinatal Society Malaysia. This is the second of four articles by him about growing up in the AI age. The information provided is for educational and communication purposes only, and should not be considered as medical advice. The Star does not give any warranty on accuracy, completeness, functionality, usefulness or other assurances as to the content appearing in this article. The Star disclaims all responsibility for any losses, damage to property or personal injury suffered directly or indirectly from reliance on such information.
